High-performance analytics for Financial Services

Challenge

Success in the financial services industry depends on the ability to apply sophisticated analytics to huge volumes of data--a feat that must be performed in ever-decreasing amounts of time. Data sources are many, spanning traditional sources like customer and transaction data and market data feeds, and emerging sources like the web and social networks. Data retention times are increasing in response to regulation and new opportunities for exploitation. And the applications that process this data are as varied as the data sources themselves, including risk, fraud and trade analytics; pricing, marketing, behavioral and sentiment analysis; and more.

Requirements

To deliver the timely analysis required to achieve and grow profits, financial services organizations require an open and flexible analytics solution, delivering the ability to:

• Capture and stage data from a variety of sources, formats and access protocols, as quickly as it arrives
• Apply sophisticated analytical techniques against this data, including transformation, indexing, MapReduce and Machine Learning
• Deliver results via existing systems such as relational databases, data warehouses or external applications, or direct to the desktop via tools like Excel, R and SAS
• Store the raw, intermediate and processed and retain as needed

Appistry Solution

Ayrris / FINANCE delivers high-speed data ingest, storage, analysis and delivery and is fully compatible with existing enterprise data sources and analysis tools.

Features

• Extreme scalability for high-speed ingest, processing and storage
• Elimination of I/O and network bottlenecks
• Powerful data management layer for existing HPC clusters
• Automated analytics pipelines
• Support for a variety of analytics algorithms such as:
  • Dense linear algebra (vector-vector; matrix -vector; matrix-matrix)
  • Sparse linear algebra
  • Spectral methods (image analysis)
  • N-body methods (motion prediction; causal models)
  • Structured grids (mesh analysis; fluid dynamics)
  • Unstructured grids (spatial analysis)
  • Monte Carlo simulations (pattern ID based on statistical sampling)
  • MapReduce
  • Combinatorial logic
  • Graph Traversal
  • Dynamic Programming
  • Back-trac/Branch & Bound
  • Graphical Model Inference (e.g., HMMs, Bayesian netowks)
  • Finite state machines
  • Statistical Computing (Regression)
• Reliable, petabyte-scale distributed storage for unstructured data
• Accepts a wide variety of data sources and formats